Is you doing git commits on a criminal fucking conspiracy?
October 9, 2023 8:56 PM   Subscribe

Samuel Bankman-Fried, founder of the cryptocurrency exchange FTX (previouslies), is on trial for fraud. His former colleagues are walking the jury through the crimes they committed in the code they committed.

More from Molly White: Mastodon ("who cares about prison time, my worst nightmare is my shitty commit messages making it into the court record"), The FTX Files.
posted by clawsoon (119 comments total) 33 users marked this as a favorite
 
Another fun snippet from this exhibit: the 5,250,000 FTT that was supposedly in the insurance fund was just hardcoded in the frontend.

ROFL. Are jurors allowed to bring popcorn?
posted by qxntpqbbbqxl at 9:08 PM on October 9, 2023 [10 favorites]


The one time I ended up with code involved in a court case it my teeth ache and that was just a question of whether or not we violated someone's very janky patent. (The lawyers kept me at arm's length because an angry engineer is not what they wanted in front of the jury).

For an actual, honest to god act of fraud? Good lord, no. I'll keep making semi-obscure commit notes!
posted by drewbage1847 at 9:17 PM on October 9, 2023 [5 favorites]


A+++ for the title.
posted by Gorgik at 9:47 PM on October 9, 2023 [34 favorites]


The idea of walking a jury through code gives me cold sweats.
posted by praemunire at 9:48 PM on October 9, 2023 [18 favorites]


We were writing an RFC while producing a reference implementation and I used the commit message "Make the structures match in case some poor bastard ever tries to implement this from the spec."

Later at the deposition it didn’t seem quite as funny.
posted by Tell Me No Lies at 9:52 PM on October 9, 2023 [29 favorites]


def _get_change() -> decimal:
...
return f2d(numpy.random.normal(7400, 3000)) * daily_volume / decimal('1e9')

*SNORT*

hahahaha

I dunno, I'd like to see the judge's reaction when someone explains random number generators to them.
posted by adept256 at 9:52 PM on October 9, 2023 [13 favorites]


I dunno, I'd like to see the judge's reaction when someone explains random number generators to them.

The judge in the case has pretty much lost his patience with the defense from day one - to the point that in a instance that Hollywood would reject for being cliched, he sustained an objection from the prosecution before it was made.
posted by NoxAeternum at 9:58 PM on October 9, 2023 [20 favorites]


That random number stuff is the world's smartest nine year old criminal's foolproof master plan.
posted by Flunkie at 10:01 PM on October 9, 2023 [8 favorites]


Sidebar: The title ought to read "Is you doing git commits..." "Is", not "Are". You drop the n-word from the original quote, fine, I get it. But don't be doing Stringer dirty by fixing up his grammar. Third person singular present tense, please! Reference: Original quote.
posted by techSupp0rt at 12:27 AM on October 10, 2023 [8 favorites]


The NumPy manual even says you're not supposed to use numpy.random.normal() for crypto!
posted by ryanrs at 12:36 AM on October 10, 2023 [29 favorites]


Pro tip: Don't use Simpsons references to name security devices, or you'll have the security footage from "Car Hole" referenced in a court like I did.
posted by krisjohn at 1:15 AM on October 10, 2023 [21 favorites]


Wow. Just... holy shit.

I haven't been following this in any detail. I mean, some big crypto thing turns out to be a scam, why am I not surprised, some dude I don't know is maybe gonna go to jail, but grifters gonna grift. You know, business as usual.

But this! It's just IN THE FUCKING CODE for ALL TO SEE! Discretion, people, learn some discretion!
posted by inexorably_forward at 1:20 AM on October 10, 2023 [5 favorites]


git --blame
posted by srboisvert at 1:58 AM on October 10, 2023 [32 favorites]


Sidebar: The title ought to read "Is you doing git commits..." "Is", not "Are". You drop the n-word from the original quote, fine, I get it. But don't be doing Stringer dirty by fixing up his grammar.
git commit -m 'Is you is or are you aint my baby'
posted by clawsoon at 3:46 AM on October 10, 2023 [11 favorites]


FTX combines the tendency of really privileged people to think their actions aren’t illegal because they want to do them with “the best child” tendency to keep records and write clean code, and also “stupid kid joke” elements like naming one of their internal channels “wire fraud.” It’s a dizzying combination of hubris, criminal intent, and lack of true capacity.

I think we’ll see a lot of “if I put this n a script, I’d have to delete it for being too on the nose” moments in the reporting.
posted by GenjiandProust at 4:36 AM on October 10, 2023 [15 favorites]


As far as I can tell, it's not even as sophisticated or opaque as some crypto thing. It looks mostly like plain old fraud, with a dash of witness tampering and campaign finance violations thrown in. The novelty mostly seems to stem from the media campaign he seemed to enjoy, and his goofy appearance.
posted by 2N2222 at 4:57 AM on October 10, 2023 [2 favorites]


Eh, the crypto aspect let them make up values. I think, in the world of real finance, people would have caught in much sooner. Without the volatility of crypto and the nonsense of “market caps,” they wouldn’t have had anywhere near the space to hide shenanigans. I agree that, like Elizabeth Holmes, personal and family connections allowed Bankman-Fried to access more clients than your average grifter.
posted by GenjiandProust at 5:37 AM on October 10, 2023 [6 favorites]


FTX combines the tendency of really privileged people to think their actions aren’t illegal because they want to do them with “the best child” tendency to keep records and write clean code, and also “stupid kid joke” elements like naming one of their internal channels “wire fraud.”

This whole trial feels like a redux of Elon Musk in the past few years. “Oh, he’s just a blithering moron.”

Maybe all this time we just weren’t dumb and amoral enough to “succeed”? Yay, I guess?
posted by leotrotsky at 6:00 AM on October 10, 2023 [8 favorites]


return f2d(numpy.random.normal(7400, 3000)) * daily_volume / decimal('1e9')

In a way it’s a shame that they had a defendant turn and reveal this to the investigators. Could you imagine being an expert witness auditing the code and coming across this line? The mother of all “What the f@ck?” moments.
posted by Tell Me No Lies at 6:10 AM on October 10, 2023 [2 favorites]


"There's no paper trail, right?"

"Um, yeah, there's definitely no paper trail."
posted by tommasz at 6:16 AM on October 10, 2023 [8 favorites]


I never had to testify, but when I worked at Enron, one of my coworkers exclaimed, "that's what they were doing!" while we were watching congressional hearings. Then a bunch of my coworkers sued a bunch of my other coworkers.
Fun times.
posted by Spike Glee at 6:29 AM on October 10, 2023 [15 favorites]


Funny that a Merkle tree is being used to take down a business based on Merkle trees.
posted by clawsoon at 6:30 AM on October 10, 2023 [9 favorites]


> Funny that a Merkle tree is being used to take down a business based on Merkle trees.

And they say Americans have no sense of irony.
posted by lhauser at 6:44 AM on October 10, 2023 [2 favorites]


You're all going nuts over the random number call, I'm completely boggled by the allow_negative flag. Holy fucking shit.
posted by JoeZydeco at 7:04 AM on October 10, 2023 [6 favorites]


The whole crypto "industry" is so awash with fraud and double dealing that an FTX-style collapse is inevitable every few years. In theory it should be possible to make a profit by running an honest exchange that charges modest fees, earns a little interest on the float, and maybe selling trade data. But such a hypothetical exchange would be immediately undercut by FTX-of-the-day offering crazy (fraudulent) deals to draw in customer money. No honest exchanges can compete so no honest exchanges exist.

This trial will not be the final blow to crypto but I hope the negative publicity helps to take some of the air out of the wider crypto markets.
posted by AndrewStephens at 7:12 AM on October 10, 2023 [7 favorites]


You're all going nuts over the random number call, I'm completely boggled by the allow_negative flag. Holy fucking shit.

Yeah, it's not as blatantly obvious as a random number call in financial code but its existence is pretty damning.
posted by Tell Me No Lies at 7:16 AM on October 10, 2023


Note to self: Hire old-school programmers for criminal conspiracies who don't believe in self-documenting variable names.
posted by clawsoon at 7:17 AM on October 10, 2023 [21 favorites]


Note to self: Hire old-school programmers for criminal conspiracies who don't believe in self-documenting variable names.

I'm still waiting for the court case where this theory gets tested. The prosecutor can connect all the dots for the jury, no matter how tangled the code is, but it's a hell of a lot tougher to convict based on do_the_thing() and do_the_thing_2() than it is on randomize_insurance_fund_balance() and const bool allow_fradulent_transactions := true;
posted by Mayor West at 8:23 AM on October 10, 2023 [9 favorites]


Myself, I don't think I could ever get into crypto scamming, because I have a longtime habit of letting my conscience get to me, and so I start my commit messages for anything really appalling the same way:

git commit -m "Forgive me for what I am about to do."
posted by Mayor West at 8:27 AM on October 10, 2023 [3 favorites]


Pro tip from someone who's worked in tech a loooong time, if someone tells you 'oh just make up a number." do not do this. It's all discoverable. If anything at all fishy looks like it is happening, make someone paid much more than you write in detail what is supposed to be happening and then file that shit away. And get your resume up to date, because you're working for at least dumb but probably also evil.
posted by Abehammerb Lincoln at 8:30 AM on October 10, 2023 [5 favorites]


If it was FORTRAN, you could at least make up some plausible-deniable backronyms, e.g.:

XFRCOS = Transfer Cash Offshore
......... or = Xtraordinary Fundraisers Reach Charitable Objectives Successfully
posted by credulous at 8:57 AM on October 10, 2023 [4 favorites]


You're all going nuts over the random number call, I'm completely boggled by the allow_negative flag. Holy fucking shit.
For me, that's because the RNG call is hysterically, outlandishly rock stupid. It's slapstick comedy - a cream pie and banana peel pratfall that harkens back to page visitor counts, but with more stakes.

The allow_negative flag is the big fraud wrapped in a simple boolean and is indicative of so much of the mind set that says "well, the computer allows me to do this, so it's ok then!"
posted by drewbage1847 at 9:08 AM on October 10, 2023 [4 favorites]


The allow negative flag is just overdraft protection. The RNG balance calculator is more "We aren't designing a trading platform, we're programming what it's interface would display in a video game."
posted by mark k at 9:24 AM on October 10, 2023 [5 favorites]


And they say Americans have no sense of irony.

Huh. I never knew this. How ironic.
posted by slogger at 9:26 AM on October 10, 2023 [2 favorites]


You're all going nuts over the random number call, I'm completely boggled by the allow_negative flag. Holy fucking shit.


I wonder if one of you super smart IT guys could give someone utterly ignorant of code and coding, which is to say me, at least a tiny hint as to what this means/ is all about? I'd be very grateful.
posted by dutchrick at 9:35 AM on October 10, 2023 [3 favorites]


The RNG bit is bad because it just means they're completely making up a fake number with no relation to reality about their insurance fund to cover bad transactions and keep the exchange liquid. It's a big lie that allows them to sell people on the idea that their crypto investments are safe, but it's all marketing crap and should never have been depended on (but people...)

The allow_negative flag allowed them to specifically exempt their internal trading accounts from restrictions placed on other accounts in terms of what they could do based on the accounts funding. (if I remember from reading last night an example would be: you can't trade if something is going to make you negative, but we can).

The way the company spun it originally was that their accounts were special for needing to do things out of bounds to cover exchange shortfalls, etc. From there, with the guard rails removed it's easy to move into pure fraud.
posted by drewbage1847 at 9:51 AM on October 10, 2023 [11 favorites]


It was basically a flag that allowed certain FTX clients to overdraft their accounts. In practice it was for one account, Alameda, a separate company also owned by SBF.

Normally, banks are prohibited from taking the money you deposit and making risky bets, such as playing the stock market, buying lottery tickets, etc. This is exactly what SBF was doing: he was taking other depositor's money and making crypto bets with it. SBF was outright lying, claiming that there was a Chinese Wall between FTX and Alameda; this code (along with testimony) proves otherwise.
posted by 1970s Antihero at 9:58 AM on October 10, 2023 [10 favorites]


mark k: The RNG balance calculator is more "We aren't designing a trading platform, we're programming what it's interface would display in a video game."

I know someone who was doing technical support for West Publishing -- home of Westlaw, "America's premiere monopolistic legal reference platform!" -- when a movie was filmed of a John Grisham novel.

For a scene where the protagonists do a bunch of Westlaw searches to Dig Up The Dirt On The Bad Guy(s), the movie-makers contacted West to get the most-realistic-possible screens of monospaced legalese, in case any attorneys went to see this legal thriller. My favorite nerd got the job to put together the text that the actors would see...and he manfully resisted the temptation to include his favorite ridiculous/outrageous cases as a distraction.

Sometimes you really do need experts behind the scenes so that your make-believe is as realistic as possible.
posted by wenestvedt at 10:01 AM on October 10, 2023 [8 favorites]


Many thank drewbage1847 and 1970s Antihero - that is very helpful. No surprise you're all wowing like crazy!
posted by dutchrick at 10:10 AM on October 10, 2023 [1 favorite]


To be fair, the coding parts of this are just extra nails in the coffin. I expect the prosecution will rely much more heavily on the other members of the fraud team testifying “oh, hell, yeah; it was fraud and we all knew it, including Bankman-Fried” and then reinforcing that with emails. Thus the defense’s desperate attempts to position Caroline Ellison as the evil mastermind, which the court seems to have little patience with.
posted by GenjiandProust at 10:13 AM on October 10, 2023 [1 favorite]


I wonder if one of you super smart IT guys could give someone utterly ignorant of code and coding, which is to say me, at least a tiny hint as to what this means/ is all about? I'd be very grateful.

(I'm sure this is a little bit off, but it gives the gist of why the code is so damning)

Part of running a brokerage is having a slush fund ("Insurance fund" in this case) so that you can settle trades immediately in case one of your clients loses so much in a day that there's going to be a delay in covering it.

Needing an enormous Insurance Fund makes it look like people are making huge trades on your system. It's PR.

FTX proudly posted their fund amount, which went a little up and down each day. That's how you could tell it was active.

Problem: They weren't actually posting the fund amount (if the fund even existed), and they were literally adding or subtracting a random number each day to make it look like it was real.

The code in question is the random number generator.
posted by Tell Me No Lies at 10:18 AM on October 10, 2023 [11 favorites]


Imagine having a gun, with smoke coming out of it, where these code comments are a tag hanging from the gun reading "murder weapon." Like, we know.
posted by wenestvedt at 10:20 AM on October 10, 2023 [4 favorites]


Big days for the “code is law” folks, maybe not in the original meaning or the Etherium usage though.
posted by theclaw at 10:26 AM on October 10, 2023 [5 favorites]


So what about the quality of the code? If it weren't for the fraud, would you hire these guys?
posted by clawsoon at 10:35 AM on October 10, 2023 [1 favorite]


It was basically a flag that allowed certain FTX clients to overdraft their accounts.

To refine on this a little, brokers trade on exchanges rather than directly in part because exchanges semi-guarantee that the trades will actually settle (that is, somebody will pay up) when the time comes (either a couple days later or whenever the agreement specifies). They arrange this by, among other things, requiring traders to post collateral with the exchange. This can be tricky if the collateral itself can vary dramatically in value over a short time, as crypto is especially prone to. If your $5 magic beans are suddenly only worth $1 on the market, then the $5 million collateral of beans you posted to the exchange are suddenly only worth $1 million. If additionally you have $3m in outstanding trades on the exchange that you can't cover (say, because the value of all your assets are collapsing and you can't get enough cash for them elsewhere), then...whoops! The exchange which could comfortably ensure that $3m in trades would be paid for out of the $5m in collateral is suddenly looking at a $2m hole.

FTX boasted about its ability to automatically close out traders' accounts before they went negative. As the value of a party's collateral dropped perilously low, FTX was supposed to liquidate its collateral, not permit further trades, etc. Not a bad idea, in principle. (This tends to be done more manually by other exchanges and with credit lines, leading to, e.g., the unbelievably hilarious situation with Archegos and Credit Suisse, where everybody said "uh, these guys should post more collateral" but no one seemed to have the courage to follow through with the client about it.) In practice, FTX (allegedly) exempted Alameda Research from this process through the allow_negative flag, basically meaning that it could lose infinite monies on collateral value of zero without being closed out. Those losses, eventually, had to be made whole, either by the exchange or (involuntarily) by the exchange's customers.
posted by praemunire at 10:56 AM on October 10, 2023 [8 favorites]


That is mind boggling, Tell Me No Lies. That sounds like shameless, fraudulent manipulation. A smoking gun as wenestvedt puts it.
posted by dutchrick at 10:57 AM on October 10, 2023


For those of us more familiar with fashion than code, I am being entertained this morning by the Guardian's typo:

"Ellison arrived in court just after 12.30pm wearing a salmon-covered dress, glasses, and a slate grey blazer."

At least, I'm assuming that's a typo and she wasn't wearing her lunch.
posted by gingerbeer at 12:00 PM on October 10, 2023 [20 favorites]


That sounds like shameless, fraudulent manipulation. A smoking gun as wenestvedt puts it.

There were so many smoking guns you could be mistaken for assuming FTX was a smoking gun factory. Dude went on a podcast and said Defi was a ponzi scheme. And that his company was basically selling ponzi shovels, like it was a good thing.
posted by pwnguin at 12:08 PM on October 10, 2023 [5 favorites]


Are they blaming Copilot yet?
posted by pmbuko at 12:15 PM on October 10, 2023 [3 favorites]


That is mind boggling, Tell Me No Lies. That sounds like shameless, fraudulent manipulation. A smoking gun as wenestvedt puts it.

Oh definitely. There is absolutely nothing subtle going on here.
posted by Tell Me No Lies at 12:46 PM on October 10, 2023


"Ellison arrived in court just after 12.30pm wearing a salmon-covered dress, glasses, and a slate grey blazer."

At least, I'm assuming that's a typo and she wasn't wearing her lunch.
You never know.
posted by mbrubeck at 12:48 PM on October 10, 2023 [8 favorites]


> The NumPy manual even says you're not supposed to use numpy.random.normal() for crypto!

Manual? I don't think I've seen that user on StackOverflow.
posted by I-Write-Essays at 12:57 PM on October 10, 2023 [2 favorites]




They're called Manuel on S-O, I-Write-Essays.
posted by k3ninho at 12:59 PM on October 10, 2023 [1 favorite]


As of now the Guardian article has been corrected to say "salmon-colored" instead of "salmon-covered".
posted by dfan at 1:04 PM on October 10, 2023 [5 favorites]


a dress even fishier than the git commits
posted by clawsoon at 1:15 PM on October 10, 2023 [7 favorites]


I… completely missed that typo in the quote in gingerbeer’s comment!
posted by eviemath at 1:15 PM on October 10, 2023 [3 favorites]


Thank you, mbrubeck, for tracking down that Ask - that's exactly what I was thinking of, of course. (Eviemath, I assumed as much!)

Now back to the smoking guns and github jokes.
posted by gingerbeer at 1:27 PM on October 10, 2023 [1 favorite]


slate grey blazer.

The one picture I saw suggested it was a knit, which...young people!!!
posted by praemunire at 2:28 PM on October 10, 2023


Many thanks to those who've explained the tech/ coding aspects of the case, an area about which I'm utterly ignorant.

Just today I was reading a really kind of heartbreaking article in the September 25th issue of the New Yorker about Bankman-Fried's parents, both Stanford Law professors, describing them as smart, high-minded, socially conscious, good people, utterly convinced of their son's innocence and apparently willing to basically bankrupt themselves for his defense -- while delicately sketching him as being completely self-absorbed and oblivious to what they're going through. Really sad.
posted by Kat Allison at 3:59 PM on October 10, 2023


Pro tip from someone who's worked in tech a loooong time, if someone tells you 'oh just make up a number." do not do this. It's all discoverable. If anything at all fishy looks like it is happening, make someone paid much more than you write in detail what is supposed to be happening and then file that shit away. And get your resume up to date, because you're working for at least dumb but probably also evil.

The company I worked for had one ginormous client for whom we handled several hundred separate accounts. But there was an economic downturn shortly after they signed up with us, and they started using our services less than we expected, so that sometimes it cost us more to provide the service than what we could legitimately bill them.

Whenever that happened the Vice President of the company would go into his office and shut the door and then open up the massive complicated spread sheet used to generate their bill and tinker with the numbers, making quite tiny changes so that the final amount billed would be sufficient for us to make a profit. The Accounts Receivable department staff didn't get asked to do that themselves, at least - they just got the edited spreadsheet back that contained the fraudulent numbers, printed out the multi-page bill and sent it out. The bills we sent them were so complicated that the client never figured out what we were doing. But the Accounts Receivable people knew, and they maintained an uncomfortable silence.
posted by Jane the Brown at 4:30 PM on October 10, 2023


Note to self: Hire old-school programmers for criminal conspiracies who don't believe in self-documenting variable names.
Oh. My. God. It's just $i all the way down.
posted by krisjohn at 4:50 PM on October 10, 2023 [8 favorites]


Oh. My. God. It's just $i all the way down.

Real Perl purists will use $_ whenever possible.
posted by notoriety public at 5:05 PM on October 10, 2023 [8 favorites]


And yet even today NPR was running an interview with a guy who wrote a book about Bankman-Fried who was still going on about what a smart guy he is and how serious he is about altruism. We desperately need tech journalists who are skeptical about tech and do their own research on tech, instead of people who breathlessly report every tech press release as fact.
posted by hydropsyche at 5:34 PM on October 10, 2023 [7 favorites]


> ... $i
         ^
Warning (line 304): Consider using a more meaningful variable name, such as $violateUsc18Section1956LaunderingOfMonetaryInstruments
posted by Flunkie at 5:36 PM on October 10, 2023 [21 favorites]


Altruism? So was he defrauding crypto bros as a charity?

...ok, I guess can see that.
posted by ryanrs at 5:53 PM on October 10, 2023 [3 favorites]


> hydropsyche: "And yet even today NPR was running an interview with a guy who wrote a book about Bankman-Fried who was still going on about what a smart guy he is and how serious he is about altruism."

Indeed. While SBF is clearly the front-runner for winning "Most Fucked" in the aftermath of this debacle, Michael Lewis appears to be jockeying for second place. Like, he could have looked around and realized that his initial, hagiographic version of SBF & FTX wasn't going to hold and gone back to his publisher and asked for time to re-write. Instead, it seems he took a look around and thought to himself "nah, SBF's fine, everyone else is just jealous and haters". It's kinda fascinating to me. Like, I wonder if Lewis is watching the trial or not because the FTX fraud seems way, way easier to understand than the mortgage crisis he wrote about in The Big Short. Tbh, between SBF/FTX and the new revelations about Michael Oher and the Tuohys (from The Blind Side), I'm not sure where Michael Lewis's reputation is going to go from here.
posted by mhum at 6:22 PM on October 10, 2023 [12 favorites]


Lewis did a neat podcast for the last few months, bringing smart folks to explain crypto stuff to him. I love his previous books and this was a great listen....but as it progressed, Lewis never actually condemned him. And the reviews of his new book from people who know crypto -- like Molly White's -- are harsh.
posted by wenestvedt at 6:43 PM on October 10, 2023 [2 favorites]


I have to imagine his choice of which narrative to present is a signal about what kind of person he thinks is likely to buy the most copies of his book. And the crypto folks have proven to be easy to influence to buy things.
posted by I-Write-Essays at 7:02 PM on October 10, 2023 [1 favorite]


I might be misremembering, but I think SBF is one of the people who define "altruism" in a very strict mathematical sense that most people would find... questionable. Or at least weird.

Like, something along the lines of "Start a colony in another galaxy" would rate higher than "Feed the hungry" because even if there's only a very small chance that our entire galaxy will be destroyed, that would be the end of humanity if we hadn't colonized another one. Potentially, bazabingabongojilliigrillions of hypothetical human lives could be saved. Non-hypothetical millions of human lives are small potatoes, even despite the extremely low odds.
posted by Flunkie at 7:17 PM on October 10, 2023 [4 favorites]


So to rephrase, his idea of altruism only cares about the 10-sigma-likelihood upside, and disregards risk entirely? I, too, remember being 14 years old.
posted by I-Write-Essays at 7:22 PM on October 10, 2023 [3 favorites]


His idea of altruism just happens to justify present day greed for speculative future goods. They call it 'effective altruism'.

Investor and entrepreneur Sam Bankman-Fried, founder of the cryptocurrency exchange FTX was involved in Effective Altruism since before 2010,[8] and around 2019 became more publicly associated with the effective altruism movement,[26] announcing that his goal was to "donate as much as [he] can".

Don't you see? He had to do as much fraud as he could so he could give away even more!

Effective altruism is psuedo-intellectual horseshit designed to soothe a greedy conscience.
posted by adept256 at 7:36 PM on October 10, 2023 [2 favorites]


They call it 'effective altruism'.

Heh. One of the groups cited as a basis for ‘effective altruism’ is Givewell. Now where have I heard that name before…
posted by Tell Me No Lies at 7:46 PM on October 10, 2023


I snagged a copy of the Michael Lewis book from the library. If it had been published before the FTX collapse, I imagine it would have been viewed as a hit job on Bankman-Fried. Despite taking SBF at his word far too often, it still paints an unflattering picture of his character and actions.

But to publish this book after the collapse and before the trial? To shrug off the question of whether the whole enterprise was a giant fraud and SBF is one of the biggest criminals in history? That is Lewis choosing the fairy tale he wanted instead of the much more interesting true story he could have gotten.
posted by mbrubeck at 7:51 PM on October 10, 2023 [2 favorites]


I had forgotten about the Givewell thing
posted by mbrubeck at 7:55 PM on October 10, 2023 [2 favorites]


I think there's a distinction, though both SBF and Givewell are shitty (and though, yeah, I think they're both referred to as "effective altruism"). Givewell was (ostensibly) trying to help people who exist and are suffering - eradicating Guinea worm, maybe? I forget, but it was something that was way less of a blatantly ridiculous application of math to charity than the stuff SBF was (if I'm remembering correctly) ostensibly into.

To be clear, my "Start a colony in another galaxy" thing was absurd hyperbole, not an actual example, but it was... absurd hyperbole of the already absurd. And yeah, I think mostly just an excuse to hoover up money.
posted by Flunkie at 7:56 PM on October 10, 2023


It just seems like a similar entitlement and dubious moral compass among the founders.
posted by Tell Me No Lies at 8:01 PM on October 10, 2023


That is Lewis choosing the fairy tale he wanted instead of the much more interesting true story he could have gotten.

On the other hand, he could’ve just been well and truly done with the book and everything else involved with it.
posted by Tell Me No Lies at 8:02 PM on October 10, 2023 [1 favorite]


return f2d(numpy.random.normal(7400, 3000)) * daily_volume / decimal('1e9')

What kind of monster uses "numpy" instead of "np"?
posted by ryoshu at 8:04 PM on October 10, 2023 [10 favorites]


Tell Me No Lies, sure, but I wasn't trying to draw a distinction between different scammy people who both base their scam in part on "effective altruism"; like I said, both SBF and Givewell suck. Rather, I was trying to say that "effective altruism" does not necessarily mean "scam" or "absurd", let alone both.

At the heart of it is just "let's use math to try to figure out what would do the most good"; that's open to problems based on ridiculous interpretations and/or lack of wisdom concerning what exactly that math should be, as well as (of course) problems based on scammers hijacking it... but neither of them is inherent in "let's use math to try to figure out what would do the most good" per se, I don't think.
posted by Flunkie at 8:12 PM on October 10, 2023


Math can give you the quantitative implications of your axioms, but it doesn't lead you to choose effective axioms. One can make logic say almost anything they wish, provided they pick the axioms that lead to the conclusions they want. That's why merely being logical is worthless.
posted by I-Write-Essays at 8:19 PM on October 10, 2023 [3 favorites]


Yes, of course, that's why I said that it's open to problems based on ridiculous interpretations and/or lack of wisdom concerning what exactly that math should be.

Unless you're seriously arguing that using math to try to help figure out how to best help is necessarily going to give a bad answer, I don't think we're disagreeing. People who describe themselves as using "effective altruism" have, for example, advocated feeding the hungry based upon it.
posted by Flunkie at 8:23 PM on October 10, 2023 [2 favorites]


I'm not disagreeing, just elaborating, because the very notion makes me angry.
posted by I-Write-Essays at 8:23 PM on October 10, 2023 [1 favorite]


> mhum: "his initial, hagiographic version of SBF & FTX"

Lemme just quickly retract this. I haven't read the book, so I have no idea if it is hagiographic or not. I think I rashly assumed that it was because I was put off by Lewis's weird tone in recent interviews.
posted by mhum at 8:33 PM on October 10, 2023


gingerbeer,

Too bad that was a typo. With an increase of over 10% in the NASDAQ Salmon Index in the past four weeks, I had hoped that Caroline Ellison had emerged from this mess with tangible assets.
posted by lukemeister at 8:36 PM on October 10, 2023 [1 favorite]


(from @mondaypunday)

Human: we have a color named after you!
Salmon: really? is it silvery blue like my outsides?
Human: no, uh–
Salmon: wait why is it pink?
Human: ...
Salmon: WHY IS IT PINK
posted by credulous at 9:23 PM on October 10, 2023 [29 favorites]


I wasn't trying to draw a distinction between different scammy people who both base their scam in part on "effective altruism"; like I said, both SBF and Givewell suck

An idea certainly isn’t responsible for the people who attach themselves to it, but what struck me is that Givewell is credited as a founding entity for effective altruism.

Given that fact, it doesn’t surprise me that the ideas appeal to people who have aspirations to help society while simultaneously abusing the trust of the people in front of them.
posted by Tell Me No Lies at 9:44 PM on October 10, 2023


I wonder if he's going to bring up the Effective Altruism stuff at his sentencing hearing, ha ha.
posted by ryanrs at 10:42 PM on October 10, 2023


In her testimony, Ellison described Bankman-Fried as "risk neutral":
She recounted him once offering up a hypothetical coin flip: If the coin turned up tails, the world would be destroyed. If it turned up heads, human life would get at least two times better. Bankman-Fried said he would take the bet.
I'm doing the math right now, and I don't think that's a neutral coin flip.
posted by clawsoon at 7:13 AM on October 11, 2023 [4 favorites]


That's funny, because just yesterday I saw an arxiv preprint of a study on fair coin flips that suggests their physical properties predispose them to land on the same side they started.

Does it destroy his entire philosophy if it turns out that fair coins aren't random? Well, maybe not, seeing as he's being tried for fraud.
posted by I-Write-Essays at 7:23 AM on October 11, 2023 [2 favorites]


Sometimes you really do need experts behind the scenes so that your make-believe is as realistic as possible.

But where could we find such a person?

This is in real time. I'll create a GUI interface using Visual Basic, see if I can track an IP address.
posted by flabdablet at 7:26 AM on October 11, 2023 [1 favorite]


A coin flip where you either double your assets or lose everything is only risk neutral in additive expectation, but in terms of expected growth rate doubling is log(2) while losing everything is log(0), so you average your small positive number with negative infinity and the expectation is not so neutral.
posted by Pyry at 7:31 AM on October 11, 2023 [11 favorites]


That's funny, because just yesterday I saw an arxiv preprint of a study on fair coin flips that suggests their physical properties predispose them to land on the same side they started.

"A group of 48 people (i.e., all but three of the co-authors) tossed coins of 46 different currencies × denominations and obtained a total number of 350,757 coin flips."

"hey guys i have an idea for a paper but you need to get your thumbs checked by your doctor first"
posted by clawsoon at 8:51 AM on October 11, 2023 [3 favorites]


"Too bad that was a typo. With an increase of over 10% in the NASDAQ Salmon Index in the past four weeks, I had hoped that Caroline Ellison had emerged from this mess with tangible assets."

Well, now that Fat Bear Week has ended, sure.
posted by gingerbeer at 9:07 AM on October 11, 2023 [2 favorites]


Rather, I was trying to say that "effective altruism" does not necessarily mean "scam" or "absurd", let alone both.

At the heart of it is just "let's use math to try to figure out what would do the most good"; that's open to problems based on ridiculous interpretations and/or lack of wisdom concerning what exactly that math should be, as well as (of course) problems based on scammers hijacking it... but neither of them is inherent in "let's use math to try to figure out what would do the most good" per se, I don't think.


If you have to modify your altruism with the adjective 'effective' that's a pretty good sign what you are doing doesn't have much face validity as altruism. Probability/Occam says you should be more suspicious of things that don't appear on the surface to be what they claim they are.

Personally, I put Effective Altruism into my Motivated Cognition - Rationalizations file.
posted by srboisvert at 9:23 AM on October 11, 2023 [2 favorites]


A coin flip where you either double your assets or lose everything is only risk neutral in additive expectation, but in terms of expected growth rate doubling is log(2) while losing everything is log(0), so you average your small positive number with negative infinity and the expectation is not so neutral.

Isn’t this basically the Martingale strategy, but for existence?
posted by atoxyl at 9:24 AM on October 11, 2023 [2 favorites]


Incidentally, here's an article from Ars Technica about Ellison's testimony, with a picture; her dress seems more rust or maroon-colored than salmon IMO. It generally seems to be a well-balanced account; of some interest is the observation that the judge was irate at the defense for repeating questions.
posted by Halloween Jack at 9:33 AM on October 11, 2023


of some interest is the observation that the judge was irate at the defense for repeating questions.

This is the behavior that resulted in the judge's preemptive sustainment I mentioned earlier. I thought "don't piss the judge off" is something that one learns as a 1L?
posted by NoxAeternum at 9:46 AM on October 11, 2023 [1 favorite]


> Isn’t this basically the Martingale strategy,

I see the similarity, but it misunderstands an important detail. The strategy doesn't work if your first bet is "All in." There's nothing left to halve when you lose on existence.
posted by I-Write-Essays at 9:48 AM on October 11, 2023 [2 favorites]


I see the similarity, but it misunderstands an important detail. The strategy doesn't work if your first bet is "All in."

You’re right, it’s actually closer to the inverse logic and as such doesn’t even make sense on the face of it. If you’d take the “double the quality of life on Earth or kill everyone” bet once, why would you not take it again - there’s not a clear point at which you’d walk away with your winnings. But if you make a strategy of repeating it the outcome converges to killing everybody. Whereas the appeal of the classic Martingale is that in the abstract the outcome of an unlimited number of flips converges to winning enough to make up for your losses - the problem is that in reality you don’t have an unlimited number of flips, eventually you are all in, which I guess is really the conceptual connection I was making.
posted by atoxyl at 11:04 AM on October 11, 2023 [2 favorites]


On the question of how much to bet when you have to bet multiple times, I found The "Just One More" Paradox interesting.
posted by clawsoon at 11:15 AM on October 11, 2023


I see it as a no-lose bet but that probably reveals me to be a psychopath
posted by some loser at 11:16 AM on October 11, 2023


I see it as a no-lose bet but that probably reveals me to be a psychopath

Or eponycontradictory.
posted by clawsoon at 11:17 AM on October 11, 2023 [2 favorites]


I see it as a no-lose bet but that probably reveals me to be a psychopath

I occasionally wonder if calling it quits as a species would be a good way to end human suffering. If that's the case then it would be a win-win.
posted by Tell Me No Lies at 12:46 PM on October 11, 2023


And in insult to injury, the thief who stole $.5B worth of crypto from FTX has been cashing it out during the trial in an impressive fuck you to the company.
posted by NoxAeternum at 12:47 PM on October 12, 2023 [1 favorite]


While a lot of the effective altruism community seems to have a bad case of engineers' disease, I think it is worth pointing out that there are some ppl in it who just take it as license to go for a high paying job and donate half their salary to a good cause. That seems like a good thing to me and not a person I'd kick out of bed for eating crackers.
posted by andorphin at 2:12 PM on October 12, 2023


I think it is worth pointing out that there are some ppl in it who just take it as license to go for a high paying job and donate half their salary to a good cause.

Given that said individual's day job is in arbitrage - that is, the sort of financial work that's been likened to a legalized bustout - this strikes me as being morally akin to CloudFlare's "hate offsets".

Or to put it another way, there's a reason you won't find many Carnegie libraries in Pennsylvania steel country, which illustrates why "effective altruism" fails as a concept.
posted by NoxAeternum at 2:25 PM on October 12, 2023 [2 favorites]


Just today I was reading a really kind of heartbreaking article in the September 25th issue of the New Yorker about Bankman-Fried's parents, both Stanford Law professors, describing them as smart, high-minded, socially conscious, good people, utterly convinced of their son's innocence and apparently willing to basically bankrupt themselves for his defense -- while delicately sketching him as being completely self-absorbed and oblivious to what they're going through. Really sad.

Sam Bankman-Fried’s Dad Thought His Son Wasn’t Paying Him Enough, So He Got Mom Involved
posted by armacy at 8:00 PM on October 12, 2023 [3 favorites]


Sam Bankman-Fried’s Dad Thought His Son Wasn’t Paying Him Enough, So He Got Mom Involved

Remind me again, which one was the ethics professor?
posted by clawsoon at 8:36 PM on October 12, 2023 [2 favorites]


This post pushed me over the edge into subscribing to Molly White's substack and it's definitely worth it if you want to mainline schadenfreude from an astute analyst who is good at explaining.
posted by i_am_joe's_spleen at 11:35 PM on October 12, 2023 [2 favorites]


Molly White is heading to New York to watch Bankman-Fried's testimony in person.
posted by clawsoon at 7:40 AM on October 26, 2023




the mortal flaw with the whole effective altruism thing, the thing that makes it something for self-regarding half-brights and people who enjoy ripping off self-regarding half-brights, is that their opening move — don't discount the future — is totally idiotic. it's the 1. e4 e5 2. ke2 of ethical systems.

like, if you don't discount the future, i.e. if you're a hammer-headed quantification-obsessed utilitarian, and also you treat hypothetical future/far-future benefits and harms as equivalent to tangible present/near-future benefits and harms rather than weighting the unpredictable future less than the actually existing now — every single decision you make involves comparing infinities to infinities. under those conditions you can come up with any nonsense plan for utility maximization you want and there's no way to really distinguish it from any other plan for utility maximization, nonsense or otherwise.

fuckin' stanford kids, fuckin' mit kids, fuckin' stanford professor parents. fuckin' self-regarding half-bright assholes.
posted by bombastic lowercase pronouncements at 8:14 AM on October 26, 2023


oh also apropos of nothing i am super long-term pissed about the press dragging caroline ellison for using millennial-speak rather than dragging her for being a cryptocurrency scammer. (that last sentence is repetitive, because "cryptocurrency" and "scam" are synonyms).
posted by bombastic lowercase pronouncements at 8:18 AM on October 26, 2023


Sam Bankman-Fried is so fucked. In her cross, she simply unhinged her jaw and ate Bankman-Fried.
posted by 1970s Antihero at 8:06 AM on October 27, 2023 [4 favorites]


From the comments on that article (great article, BTW), I learned that Molly White did a pair of livestreams about the testimony. Haven't watched them yet myself, but: part 1, part 2
posted by clawsoon at 8:30 AM on October 27, 2023 [2 favorites]


Cross has begun, and Bankman-Fried has come to a battle of wits unarmed.
posted by NoxAeternum at 5:40 PM on October 30, 2023 [4 favorites]


Cross has begun, and Bankman-Fried has come to a battle of wits unarmed.

These article from Elizabeth Lopatto at The Verge have been wonderfully snarky
posted by clawsoon at 3:04 AM on October 31, 2023 [1 favorite]




« Older Airlines Are Just Banks Now   |   It's IRC, but you get to be a cat! Newer »


This thread has been archived and is closed to new comments